Rebuilt AI
Rebuilt AI
I've been very interested in designing a highly interactive, humanistic, and hopefully even unique(to each individual player) AI. I've already gone through the planning/research process, and feel that AI design is WAY behind the technology available(as opposed to graphics and physics which have been almost fully developed to the point that our technology limits us to). Anyway, I know this would be ALOT easier designing my AI in half-life or Unreal, but I've spent literally months of my life playing halo 2 live, and understand the human decision process, and factors that are most important, and want to work with something I'm passionate about. Anyway, I need the help of the community to start creating this, as I've hit a wall. **1)Guerilla is nice and pretty thourough about editing individual AI's, but still limits what I want to do.(it allows you to change charecteristics, but those just edit values that I want to get to directly) If Guerilla is a UI to make it easier to change the design of the AI, how do I go directly to the source? I want to redefine the finite state, pathfinding used, and then implement my design to make it "learn" mistakes made. Any other way to change the AI outside of Guerrila? I've got HaloPC, CE, Xbox 1&2 on my computer, and would rather edit the halo 2 ai, but CE might present easier solutions. Anyway, I am a complete noob, and started learning halo modding last week, so any help in the right direction would be great
-
WaywornMmmmm
- Posts: 1341
- Joined: Sat Nov 05, 2005 5:17 am
- Location: U.S.A
|
You're right man, my brains honestly overloaded right now trying to seperate all the halos with thier respective programs and the functionality associated with each, as well as what I'm able to do. So how do I edit AI design in halo 2? Ideally I'd like to rebuild the H2 AI, but if that's not possible yet, I suppose I'll give it a try in CE. Any suggestions?
Last edited by nkhende on Wed Aug 30, 2006 2:14 pm, edited 1 time in total.
Not much is known about that, but I believe its controled by scripts, which have not been delt with as of now, I don't know that much so you may want to check out the AI research project here.
My real name is a Mammoth, if someone who cares sees this, please change my name to a Mammoth
I read that forum and tried to get involved, and project AI was definitely a good first step, but it seemed that the main goal was to simply introduce an already poorly designed AI into a MP map(which I'm not ripping on, I understand what a big achievement that was), and when they succesfully did that they quit. (Resurrect project AI with new goals?)
Here's why I want to do this:
1)With the speed at which computation operations are carried out, there should be no reason that AI's are as stupid and simple as they are now.
*Imagine playing any matchmade game at 1/100 the speed of halo, while others comprehended it at a normal pace.(I realize the game would take 100 times longer then everyone else, but that's irrelevant) With time slowed down that much, I beleive there is no excuse to ever make a bad decision(i.e. human error)) And with more time, the BEST decision would be made since you've had 100 times more time to think about responses(note that this still doesn't account for unknown factors). THIS is how AI's should be built, and how fast the computational power is able to "think" at. I'm not going into theory, since that doesn't belong on this board, or my design since until I have it in effect it too is just theory. I'm mearly looking for the correct path to take in editing the halo AI at it's core.
2)Consider this: With all the RSS/previous game data available(I personally have 32 pages w/ 25 games on each) for each player, individual tendencies and skill could easily be applied to a sound AI model to produce an extremely unique AI that functions like you would. Not to mention that for each game is a map saved with the exact locations of where a player dies, where a player kills another from, and what weapon achieved this. Extract those values into a database and the pathfinding algorithm would essentially write itself, and would be humanistic because it "learned" from human error.(and would be far superior to "designed" pathfinding methods)
I'm done ranting. I want to build an AI like the world has never seen before, and I know how it's going to be done, just not exactly what tools I'm going to need to learn how to use. Hell, a week and a half ago I'd never even thought about AI before until I had this stupid stoned idea, and now I can't stop noticing how wrong AI design now is
Here's why I want to do this:
1)With the speed at which computation operations are carried out, there should be no reason that AI's are as stupid and simple as they are now.
*Imagine playing any matchmade game at 1/100 the speed of halo, while others comprehended it at a normal pace.(I realize the game would take 100 times longer then everyone else, but that's irrelevant) With time slowed down that much, I beleive there is no excuse to ever make a bad decision(i.e. human error)) And with more time, the BEST decision would be made since you've had 100 times more time to think about responses(note that this still doesn't account for unknown factors). THIS is how AI's should be built, and how fast the computational power is able to "think" at. I'm not going into theory, since that doesn't belong on this board, or my design since until I have it in effect it too is just theory. I'm mearly looking for the correct path to take in editing the halo AI at it's core.
2)Consider this: With all the RSS/previous game data available(I personally have 32 pages w/ 25 games on each) for each player, individual tendencies and skill could easily be applied to a sound AI model to produce an extremely unique AI that functions like you would. Not to mention that for each game is a map saved with the exact locations of where a player dies, where a player kills another from, and what weapon achieved this. Extract those values into a database and the pathfinding algorithm would essentially write itself, and would be humanistic because it "learned" from human error.(and would be far superior to "designed" pathfinding methods)
I'm done ranting. I want to build an AI like the world has never seen before, and I know how it's going to be done, just not exactly what tools I'm going to need to learn how to use. Hell, a week and a half ago I'd never even thought about AI before until I had this stupid stoned idea, and now I can't stop noticing how wrong AI design now is
-
xXxCocoFangxXx
- Posts: 3756
- Joined: Thu Sep 08, 2005 9:57 am
- Contact:
